Anthropic releases Claude Fable 5, a generally available Mythos-class model with 1M-token context and safeguards on frontier LLM research requests

Anthropic announced Claude Fable 5 as its first generally available Mythos-class model, stating it is the same underlying model as Claude Mythos 5 with added safeguards. Fable 5 achieved a 29.3% score on FrontierCode Diamond (compared with 13.4% for the previous benchmark best) and ranked #1 on Artificial Analysis Intelligence Index at 64.9.

Fable 5 implements two controversial restrictions: (1) 30-day data retention for all traffic on Mythos-class models, with data not used to train new Claude models; (2) RSI (Responsible Scaling Initiative) safeguards that limit effectiveness on requests targeting frontier LLM development (e.g., pretraining pipelines, distributed training infrastructure, ML accelerator design) through prompt modification, steering vectors, or parameter-efficient fine-tuning (PEFT), affecting an estimated ~0.03% of traffic concentrated in fewer than 0.1% of organizations.

API pricing is $10 / million input tokens and $50 / million output tokens, with cache pricing at $12.50 / million cache writes and $1 / million cache reads. Fable 5 maintains a 1M-token context window. Anthropic initially made Fable 5 available through Pro, Max, Team, and Enterprise plans until June 22, then switched to usage credits due to capacity constraints, with plans to restore broader subscription access later.